Detection of foreign fibers in cotton on the basis of wavelet

Foreign fibers in cotton have seriously effect on cotton production the quality of cotton products, while image processing algorithms based on machine vision offer an effective measure to deal with the problem. Wavelet is introduced to detecting foreign fibers in cotton as its great potential and excellent feature in signal and image processing. In this paper, the contaminants recognition process is divided into three steps, namely transformation of picture, image processing based on wavelet and image post-processing. On the first step, the format of color image is converted to index from RGB, which is the precondition of wavelet image analysis. A piecewise linear transform model is proposed to enhance the image. Secondly, the best tree analysis structure is calculated contrasting to the initial one. Through optimizing the entropy value, wavelet packet 2-D compresses off about 90% of the original coefficients by deleting the redundant information, while saving 99.13% energy. And image is reconstituted basis on the adjusted wavelet coefficients. In the last part, wiener adaptive filter is employed to smoothing the picture, and then we dichotomize the images after such series of treatments. The effect of the image processing indicated that wavelet packet is an effective method in the foreign fibers inspection combined with some other measures.